DOI: 10.5593/sgem2017/23/S10.002


M. Mleczko, M. Mroz, P. Sliesinski, M. Fitrzyk
Monday 11 September 2017 by Libadmin2017

References: 17th International Multidisciplinary Scientific GeoConference SGEM 2017, www.sgem.org, SGEM2017 Conference Proceedings, ISBN 978-619-7408-03-4 / ISSN 1314-2704, 29 June - 5 July, 2017, Vol. 17, Issue 23, 11-18 pp, DOI: 10.5593/sgem2017/23/S10.002


The new era of SAR exploitation started with the launch of Sentinel-1A and 1B satellites providing dual-polarized VV/VH imagery, free of charge, with unprecedented revisit interval of a few days over very large swaths. The research questions addressed in this paper are: how to process efficiently big set of multitemporal Sentinel-1 images in order to retrieve correctly classified agricultural crops for inventory purposes? How to reduce the amount of data and the redundancy of information resulting from very frequent observations? Two methods of feature space reduction and decorrelation have been proposed: PCA and ICT. The classification procedure was based on the maximum likelihood classifier. The results show that combining the ascending and descending orbits is necessary in order to avoid misclassification of crops characterized by row tillage. On the other hand the combining of VV and VH polarizations increased the accuracy of separation of potatoes and maize. Winter cereals and summer crops other than potatoes and maize were confused at about 38 percent. The rapeseed is successfully classified with very high accuracy on short time series of images at VV and VH polarizations.

Keywords: crops mapping, Sentinel-1, C-band, dual-polarimetric images, ICT/PCA